Computation of approximate entropy and sample entropy for characterization of respiratory diseases

Abstract

Capnography is a non-invasive method that provides useful information for assessing respiratory diseases. The analysis of the capnogram waveform is frequently based on the exploration of capnogram indices. However, those indices are not clearly identified when there are respiratory abnormalities. In this study, entropy measures, specifically approximate entropy and sample entropy were proposed as features for the analysis of regularity of capnogram waveforms. The study was conducted on capnogram recordings collected from asthma and pulmonary edema patients. The results showed that pulmonary edema demonstrated higher approximate and sample entropy values than asthma, which can reflect higher irregularities in pulmonary edema capnograms. Moreover, the effect of varying input parameters on entropy measures was explored, and a greater effect was found for the approximate entropy algorithm. A capnogram segment recorded for 60 seconds was suggested as a suitable length for regularity analysis in a capnogram waveform.